Rear side dielectrics on interdigitating p-(i)-n back-contact solar cells hydrogenation vs. charge effects

Polysilicon-on-oxide (POLO) passivating contacts and interdigitated back-contact (IBC) cell technologies have recently attracted a lot of interest as candidates for the implementation in the next generation of solar cells. An IBC cell with POLO junctions for both polarities a POLO-IBC cell has to electrically isolate the highly defective p and n poly-Si regions on the rear side of the cell to avoid parasitic recombination. Inserting an initially undoped, intrinsic (i) region between the p and n poly-Si regions was demonstrated to successfully prevent the parasitic recombination in the transition region of ISFH’s 26.1%-efficient POLO-IBC cell. In order to further improve the conversion efficiency towards 27%, we apply hydrogen-donating dielectric layer stacks to the p-(i)-n POLO interdigitating rear side to enhance the passivation quality of the POLO junctions. We indeed show a significant improvement of POLO junctions on symmetrical full-area homogenously doped reference samples, but when we apply a hydrogen-donating layer stack on the p-(i)-n POLO interdigitating rear side, we observe a strong degradation in the performance of the POLO-IBC cell. We attribute this to the formation of a conductive channel between the p and n poly-Si regions due to the strong negative charge density of the hydrogendonating layer stack.


Introduction
The photovoltaic (PV) market demands for ever higher PV module efficiencies. Polysilicon-on-oxide (POLO) passivating contacts and interdigitated back-contact (IBC) cell technologies have recently attracted a lot of interest as candidates for the implementation in industrial production in the near future. However, the realization of an IBC cell with POLO junctions for both polarities À a POLO 2 -IBC cell À requires a separation of the highly defective p + and n + poly-Si regions on the rear side of the cell to avoid parasitic recombination. Beside the trench separation [1] and local oxidation of the poly-Si [2], inserting an initially undoped, intrinsic (i) region between the p + and n + poly-Si regions À as reported by several research groups [3][4][5][6] À was demonstrated to successfully prevent the parasitic recombination in the transition region [5,7]. In 2018, a POLO 2 -IBC cell with such a p + -(i)-n + POLO interdigitated rear side achieved an efficiency of 26.1% [5], even without an optimized hydrogenation scheme for the POLO junctions. Thus, we recently applied hydrogendonating layer stacks on symmetric POLO junction samples to demonstrate a significant improvement of the passivation quality compared to that implemented in the 26.1%-efficient cell from 2018 [8]. Moreover, we developed the required laser ablation process for the hydrogendonating layer stack to be able to create laser contact openings without damaging the POLO junction underneath [9]. In this contribution, we study the interplay of hydrogen-donating layers with the p + -(i)-n + POLO interdigitated rear side of our POLO 2 -IBC cell. Since the typical hydrogen-donating layers like Al 2 O 3 or SiN y accommodate a high positive or negative charge, we investigate the influence of such a charge on the POLO 2 -IBC cell performance. We find that any strong charge density at the p + -(i)-n + POLO interdigitated rear side leads to enhanced non-ideal recombination and diminishes the performance of POLO 2 -IBC solar cells.

Experimental
2.1 POLO 2 -IBC solar cell precursor with different rear side dielectric layers Figure 1a shows the structure of the precursor stage of our POLO 2 -IBC solar cell fabricated similar to references [5,7]. One characteristic feature of this cell is the p + -(i)-n + POLO interdigitated rear side, where the p + -type poly-Si and n + -type poly-Si regions are separated by an initially intrinsic (i) poly-Si region. The term "initially" refers to the fact that during high-temperature annealing, strong lateral diffusion of dopants in the poly-Si (on the length scale of several micrometres) yields an inter-diffused lateral junction [7].
In brief, the following fabrication process yielded the cell precursor in Figure 1a: A 2.2 nm-thin interfacial oxide layer was thermally grown onto a 1.3 Vcm, p-type FZ wafer, which was subsequently capped by a low pressure chemical vapor deposited (LPCVD) intrinsic amorphous Si layer. We performed masked ion implantation of boron and phosphorous into the amorphous Si on the rear side and a blanket phosphorous implantation on the front side. For the former, the PECVD-deposited SiO x implant mask barriers were patterned by photolithography, such that a gap of 30 mm i poly-Si remained between the n + -type and p + -type poly-Si fingers. Then, the amorphous Si recrystalized during a pyrogenic oxidation at 900°C to grow a ∼200-nm-thick SiO 2 layer and the POLO junction formed in a subsequent higher temperature process at 1035°C. This step also yielded a lateral interdiffusion of dopants on the length scale of several micrometers, resulting in an inter-diffused junction within the initially 30 mm wide intrinsic poly-Si region [7]. The inter-diffused junction still isolates the charge carrier in the n + poly-Si from that in the p + poly-Si region. Hartenstein et al. have recently confirmed experimentally that the compensation of dopants at the p + -(i)-n + poly-Si junction provides a highly resistive narrow region, which separates n + poly-Si and p + poly-Si electrically [10].
As in references [5,7], the POLO junctions on the rear side were hydrogenated by depositing a sacrificial hydrogen-containing, silicon-rich SiN y layer on the thick thermal SiO 2 layer and by subsequently annealing for 30 min at 425°C. Eventually, the SiN y layer was removed in hot phosphoric acid.
Starting with a cell precursor as in Figure 1a, which has resulted in a cell efficiency of 26.1% in reference [3], an advanced hydrogenation scheme from Section 2.2 is applied to obtain the cell precursor in Figure 1b with improved POLO junctions on the rear side. For this purpose, we removed the SiO 2 layer from the front side and texturized it. Then, we remove the SiO 2 layer from the rear side and passivated the front side. On the rear side, we deposited a layer stack of a 20-nm-thick Al 2 O 3 , a 30-nm-thick hydrogen-containing silicon-rich SiN y and a 200-nm-thick PECVD-SiO z , which was subsequently annealed for 30 min at 425°C.
Alternatively, we removed the layer stack from the rear side of the cell precursor in Figure 1b by a single-sided HF etching and replaced it by a 20-nm-thick PECVD-SiO z and a 30-nm-thick SiN y to obtain the cell precursor in Figure 1c. As indicated in Figure 1c, the 200-nm-thick SiO z is missing on the rear side of the sample for simplicity, but has to be included in the final cell for optical reasons. The minority carrier lifetime of each cell precursor is characterized by spatially resolved infra-red lifetime mapping (ILM) [11].
Furthermore, using the same front-end process of the cell precursors in Figure 1a and b, symmetrical reference wafers with full-area doped POLO junction are fabricated and characterized by using the Sinton lifetime tester.

Advanced hydrogenation of POLO junctions
While the ∼200 nm-thick thermally grown SiO 2 in Figure 1a was necessary in reference [5] to enable a damage-free local laser ablation to create contact openings, it hinders the hydrogenation of the POLO junctions from hydrogen-donating SiN y layer to some extent. We replaced the SiO 2 by a layer stack of Al 2 O 3 /hydrogen-rich SiN y / Al 2 O 3 , which is known to be an efficient hydrogen-donating layer stack [12] to estimate the upper limited for the passivation quality of the POLO junctions used for the 26.1%-efficient POLO 2 -IBC cell. To study the effect of the improved hydrogenation scheme, symmetric p + POLO, n + POLO and iPOLO samples were prepared following the procedure in Section 2.1. After POLO junction formation, the thermally grown SiO 2 was replaced by Al 2 O 3 / hydrogen-rich SiN y /Al 2 O 3 and annealed at temperatures above 425°C.

Charging of the p + -(i)-n + rear side of the POLO 2 -IBC solar cell precursor
In order to investigate the influence of charged dielectric layers at the p + -(i)-n + rear side on the cell performance, we systematically manipulated the charge density at the rear side of cell precursor in Figure 1a and monitored the recombination behavior of the cell precursor at each charging condition. For this purpose, we deposited corona charges on top of the thermally grown SiO 2 by means of a needle-plate electrode corona discharge at 8 kV. A single charging step yielded a charge density of 4.5e11 cm À2 . We measured the recombination behavior of the cell precursor by using infrared lifetime mapping [11] at different illumination intensities and calculated the illumination-dependent implied open-circuit voltage Suns-iV OC characteristic of the cell precursor at each charging state. Figure 1d shows the lifetime map of the cell wafer from Figure 1a, which contains seven cell regions (a-g) and four reference regions for iPOLO. The lifetime within the cell region b and e reach values of 1.8 ms due to the good passivation of the n + POLO and p + POLO junctions. In this stage, the full-area POLO references reveal a saturation current density of 4 fA/cm 2 and 5.5 fA/cm 2 for n + POLO and p + POLO junctions, respectively.

Results
However, as already reported in reference [7], the intrinsic reference region (iPOLO) in Figure 1d is almost unpassivated and therefore yields low lifetimes. In contrast to that, properly hydrogenated iPOLO junctions have been reported elsewhere [4]. This is a strong evidence that the thick SiO 2 hinders hydrogen from reaching the thin interfacial SiO z of the iPOLO junctions of the 26.1%-efficient cell in reference [5] and that the hydrogenation of n + and p + POLO junctions is not optimal. Thus, it seems straight forward to apply a better hydrogenation scheme to improve the POLO 2 -IBC cell efficiency of 26.1% towards 27%.
Indeed, when replacing the thick SiO 2 layer by the Al 2 O 3 /hydrogen-rich SiN y /Al 2 O 3 layer stack and performing an annealing above 425°C, we have been able to improve the saturation current density of our n + POLO and p + POLO junctions on symmetric lifetime samples down to below 0.5 ± 03 fA/cm 2 and 3.3 ± 0.7 fA/cm 2 [8], respectively. For the iPOLO junction sample, an effective surface recombination velocity S eff of approximately 12 cm/s is achieved after hydrogenation, which is more than an order of magnitude lower compared to the iPOLO junction of the 26.1%-efficient cell [7]. The improved hydrogenation of the n + and p + POLO junctions from 4 fA/cm 2 and 10 fA/cm 2 to 0.5 fA/cm 2 and 3.3 fA/cm 2 was estimated to result in an efficiency improvement of ∼0.4% abs. by means of device simulations [8].
To prove the benefit of an improved hydrogenation scheme on the cell level, we deposited a similar hydrogendonating Al 2 O 3 /SiN y /SiO z layer stack as evaluated above on the cell precursor. This results in the structure shown in Figure 1b. Figure 1e depicts the lifetime map of the cell precursor. As expected from the hydrogenation experiments above, the iPOLO reference region of the wafer improves significantly upon hydrogenation and exhibits a lifetime of 1.5 ms. After hydrogenation, the full-area POLO references of the cell precursor in Figure 1b reveal a saturation current density of 2.5 fA/cm 2 and 4 fA/cm 2 for n + POLO and p + POLO junctions, respectively. Regardless of the improvement of the POLO junctions, the lifetime within cell regions drops from previously 1.5-1.8 ms for the cell precursor as shown in Figure 1a to about 200-300 ms for that in Figure 1b.
We explain this surprising observation by the high density of negative fixed charge in the Al 2 O 3 layer in contrast to the almost neutral SiO 2 layer. The negative charge density in the dielectric layer or at its interface to poly-Si causes an upward band bending within the poly-Si layer close to its interface with the dielectric layer [13]. Thus, we expect an accumulation of holes and a depletion of electrons in the first few nm of the poly-Si layer at the Al 2 O 3 interface in the degenerately doped p + and n + poly-Si region, respectively.
In the initially undoped region between the p + and n + poly-Si regions, the compensation of dopants after high temperature annealing results in a moderately doped, graded pn junction. In the case of an almost absent surface charge on top of the poly-Si layer e.g., with a SiO 2 layer, one part of this transition region exhibits a high resistivity and ensures electrical isolation between the p + and n + poly-Si regions. However, upon applying the Al 2 O 3 layer, its strong negative fixed charge manipulates the carrier population in the moderately doped transition region, such that an accumulation and inversion layer form in the moderately doped p-type and n-type poly-Si regions, respectively. The carrier population modulation and the transport barrier lowering due to electrostatics significantly increase the carrier mobility within the poly-Si [14][15][16][17][18]. Thus, the Al 2 O 3 effectively establishes a hole-conductive channel at its interface to the poly-Si at the p + -(i)-n + POLO interdigitated rear side [13]. This conductive channel connects the highly defective p + and n + poly-Si regions and results in a strong recombination of electrons with holes. A similar enhanced recombination was already observed for POLO 2 -IBC cells without an initially intrinsic poly-Si between p + and n + poly-Si regions [6,19,20].
In order to confirm our hypothesis, we corona charge the SiO 2 surface of the cell precursor in Figure 1a   illumination intensity characteristic. We systematically manipulate the charge density on the rear side by depositing a corona charge density Q C on top of the SiO 2 . First, we investigate the effect of negative charges and notice that the effect of charging the rear side on the lifetime of the precursor weakens over time. We speculate that this second order effect is due to the instability of corona charges on top of the SiO 2 layer [21,22]. On the time scale of the experiment, this instability causes an uncertainty with respect to the actual amount of charges present on the SiO 2 . Qualitatively, the experiment still yields valid data. Moreover, we utilize this effect to return to the state of the cell precursor without corona charging. To reach this state, we deposit approximately the same amount of positive charges as negative charges previously applied and wait about 60 hours for the charge to neutralize. We start from this state of the precursor to study the effect of positive corona charges applied to the SiO 2 . Figure 2 summarizes the Suns-iV OC characteristics for different Q C values. If negative corona charges are deposited, the Suns-iV OC curve at implied voltages close to the maximum power point of the final 26.1% cell (640 mV [5,7]) shifts towards higher illumination intensities, which corresponds to an increased non-ideal recombination with a high local ideality factor (up to 3) and which diminishes the implied pseudo fill factor. Close to one-sun conditions, the effect of charging is minor and the local ideality factor is about 0.9 for the uncharged state and 1.2 with a corona charge density of À2.7e12 cm À2 . For positive corona charges, the Suns-iV OC characteristic improves slightly up to a corona charge density of 0.9e12 cm À2 . Further increasing the positive charge densities lead again to enhanced non-ideal recombination at around a voltage of 640 mV À similar to negative charges. Figure 3 shows the implied pseudo-efficiency ih of the cell precursor, which results from the Suns-iV OC characteristic with an assumed constant short-circuit current density of 42.6 mA/cm 2 as has been measured for the 26.1% efficient cell in references [5,7]. Without corona charging, the cell precursor has an ih of 26.4%, which drops suddenly for negative Q C . If a corona charge density of 0.5e12 cm À2 or 0.9e12 cm À2 is applied, the ih increases to about 26.5%, but drops again sharply for larger Q C .
Given the results in Figures 2 and 3, it can be understood why the cell precursor in Figure 1b with Al 2 O 3 on the rear side with a typical charge density of À4e12 cm À2 and the cells in reference [6] exhibits a worse recombination behavior and a compromised cell performance À even if the POLO junction's passivation quality is improved upon hydrogenation. Moreover, we expect that a cell with a strongly positively charged SiN y hydrogen-donating layer would cause the same effect. Therefore, an IBC cell with a p + -(i)-n + POLO interdigitated rear side requires an almost uncharged hydrogen-donating dielectric layer on the rear side À at least in the transition region of the p + -(i)-n + junction.
Replacing the strongly charged Al 2 O 3 layer by a weakly charged À optimally hydrogen-containing À PECVD/ ALD-SiO z or SiO z N y on the rear side of the cell precursor in   1b poses one possibility for such a dielectric layer stack, which could also be compatible with the laser contact opening process in reference [9].
To demonstrate that a PECVD-SiO z meets the requirement of being weakly charged and that it is suitable for the application on the rear side of POLO 2 -IBC cells, we removed the Al 2 O 3 /SiN y /SiO z dielectric layer stack from a cell precursor as shown in Figure 1b and subsequently deposited a non-optimized PECVD-SiO z and a SiN y on the rear side. The resulting cell structure is depicted in Figure 1c. The cell precursor with a PECVD-SiO z /SiN y layer stack on the rear side shows a strongly improved performance in Figure 1f compared to that with an Al 2 O 3 / SiN y /SiO z layer stack in Figure 1e. The cell precursors in Figure 1f exhibit an implied pseudo-efficiency of up to 26.3%. This is qualitatively comparable with the cell precursor from Figure 1d without corona charging. However, the front side passivation of the cell precursor in Figure 1a deviates from that in Figure 1c and the meaningfulness of a comparison of both cell precursors on an absolute efficiency scale is limited. However, the successful implementation of the PECVD-SiO z on the rear side of a POLO 2 -IBC cell precursor represents a good starting point for future optimization towards 27%efficient POLO 2 -IBC cells.

Conclusion
In this contribution, we aimed at implementing hydrogendonating layers on the p + -(i)-n + POLO interdigitated rear side to improve our POLO 2 -IBC cell with an efficiency of 26.1% towards an efficiency of 27%. Since the typical hydrogen-donating layers like Al 2 O 3 or SiN y accommodate a high positive or negative charge, we investigated the influence of such a charge on the POLO 2 -IBC cell performance. We fabricated POLO 2 -IBC cell precursor with three different dielectric layers/layer stacks on the rear side: (i) a thermally grown SiO 2 as has been used for the 26.1% cell; (ii) a hydrogen-donating Al 2 O 3 /SiN y /SiO z stack to improve POLO junction passivation quality; (iii) a hydrogen-donating PECVD-SiO z /SiN y stack. The implied pseudo-efficiency ih of each precursor was deduced from the illumination-dependent implied open-circuit voltage Suns-iV OC of the cell precursor.
While the cell with thermally grown SiO 2 showed the expected ih of 26.4%, a comparison with the second cell precursor indicated that the hydrogenation scheme with the thick SiO 2 is less efficient compared to the Al 2 O 3 /SiN y / SiO z stack. The second cell precursor with Al 2 O 3 /SiN y / SiO z stack efficiently hydrogenated the POLO junction, but yielded a cell precursor with worse recombination behavior and low performance. We attributed this surprising observation to the presence of a strong negative charge induced by the Al 2 O 3 layer.
To confirm this hypothesis, we studied the influence of charges at the p + -(i)-n + POLO interdigitated rear side on the POLO 2 -IBC cell performance. We systematically manipulated the charge density by depositing a corona charge density on top of the SiO 2 covering the rear side. We found that strong positive or negative charge densities cause an enhanced non-ideal recombination and degrade the cell performance significantly, which is most likely due to the formation of a conductive path for electrons or holes within the highly defective poly-Si. Fortunately, the thermally grown SiO 2 has a nearly ideal charge density and its ih can only slightly be improved by 0.1% abs to 26.5% by adding a positive corona charge density of below 0.9e12 cm À2 .
We conclude that it is vital to control the charge density of the dielectric layers applied to the p + -(i)-n + POLO interdigitated rear side to avoid a strong increase of non-ideal recombination at the p + -(i)-n + poly-Si junction. While this effect is highly pronounced for IBC cells with poly-Si, it may also apply À to some extend À to all types of IBC cells with charged dielectric layers on the interdigitated rear side.
Finally, we presented a weakly charged hydrogendonating layer stack comprising a 20-nm-thin and a potentially weakly charged PECVD-SiO z at the interface with the poly-Si capped by a hydrogen-donating SiN y layer. This stack may provide an efficient hydrogenation scheme, while maintaining the isolation between p + and n + poly-Si regions. A first cell precursor with the weakly charged hydrogen-donating layer stack yielded an ih of 26.3%.